About

Tim Verbelen, Senior Researcher at Ghent University - imec

Tim Verbelen

Tim Verbelen received his M.Sc. degree in Computer Science from Ghent University, Belgium in June 2009. In July 2013, he received his Ph.D. degree with his dissertation "Adaptive Offloading and Configuration of Resource Intensive Mobile Applications". In October 2014, he joined iMinds IoT lab working on distributed intelligence for supporting the next generation IoT applications. As of October 2016, he works as senior researcher at imec on deep learning in distributed and resource-constrained environments.

Since 2013, Tim is a committer of the Eclipse Foundation, contributing to the Concierge project, a lightweight OSGi core implementation. As of October 2014, he is also active as Invited Researcher of the OSGi Alliance.

For a list of my publications, please take a look at Google Scholar

Research


My research focuses on Distributed Intelligence: creating smart applications in a distributed environment.

DIANNE: DIstributed Artificial Neural NEtworks

DIANNE is a modular software framework for designing, evaluating and training artificial neural networks. It is build on top of OSGi and the open source AIOLOS platform, which allows to transparently deploy and redeploy (parts of) a neural network on multiple devices. DIANNE also supports distributed training of neural networks, both supervised as in a reinforcement learning context.

DIANNE project site

AIOLOS: mobile code offloading

The AIOLOS middleware framework allows to migrate parts of the application at runtime to infrastructure available in the network. This way, resource-intensive parts of the application can be offloaded to more powerful resources in the network. A central concept in the AIOLOS framework is application modularity: the application is built up from loosely coupled software components. To do this, we heavily rely on the OSGi specification, the de facto standard for software modularity in Java.

Verbelen, Tim, Pieter Simoens, Filip De Turck, and Bart Dhoedt. 2012. "AIOLOS: Middleware for Improving Mobile Application Performance Through Cyber Foraging." Journal of Systems and Software 85 (11): 2629-2639.

Cloudlets: computing at the edge of the Cloud

One of the biggest drawbacks of a centralized back-end cloud datacenter, is the fact that this introduces a high latency link to the Cloud and the network becomes the bottleneck. This may be unsuitable for (soft) real-time applications where the latency should be kept within certain bounds. To address this issue, we envision a deployment of smaller, but more geographically distributed infrastructure that can be used for processing, so called Cloudlets. The use of Cloudlets can solve (or mitigate) many of the drawbacks of a centralized infrastructure, but at the cost of more complex management.

Verbelen, Tim, Pieter Simoens, Filip De Turck, and Bart Dhoedt. 2013. "Leveraging Cloudlets for Immersive Collaborative Applications." IEEE Pervasive Computing 12 (4): 30-38.

Software partitioning algorithms

Suppose a highly distributed infrastructure is available, and everyone wants to use this infrastructure to run distributed mobile applications on, the question remains which parts of which application of which user should run on which server. This is a huge optimization problem to solve, and heuristics are required to find a (good) solution to this problem in a timely manner. We research different algorithms to address this problem, and evaluate them based on solution quality and speed.

Verbelen, Tim, Tim Stevens, Filip De Turck, and Bart Dhoedt. 2013. "Graph Partitioning Algorithms for Optimizing Software Deployment in Mobile Cloud Computing." Future Generation Computer Systems 29 (2): 451-459.

Concierge: a lightweight OSGi runtime for mobile and embedded devices

In our research prototypes, we often use the OSGi specification for modular Java applications. Therefore, we are also involved in the Concierge project, a small-footprint implementation of the OSGi Core Specifications optimized for mobile and embedded devices. This project is currently hosted by the Eclipse foundation.

Concierge Eclipse Project

Dissertation

Adaptive Offloading and Configuration of Resource Intensive Mobile Applications

Due to the increasing popularity of mobile devices such as smartphones and tablets, more and more effort is spent on mobile application development. The development of mobile applications rises additional software engineering challenges. First, one should take into account the limited processing power and battery life of mobile devices. Second, the application should run on the myriad of different devices that is available on the market today, each with its own hardware specifications. Third, due to the fact that the user is mobile, the application can also be used in a huge number of different contexts, such as at home, at work or on the train, which affects various context parameters, such as for example the network connectivity.

To address these challenges, this dissertation presents an autonomic middleware framework, which facilitates mobile application development by automatically adapting mobile applications to the capabilities of the mobile device and the given context. For example, parts of the application can be outsourced to fixed infrastructure in the network, or the quality of the application can be adjusted depending on the available processing power of the device. Several heuristic algorithms have been proposed and evaluated to solve this optimization problem. Also, a number of compute-intensive mobile multimedia applications have been developed in the domain of image processing and Augmented Reality, in order to demonstrate the applicability of the proposed techniques.

[Download]

Contact

Tim Verbelen
tim.verbelen_at_ugent.be
Technologiepark 15
9052 Zwijnaarde
Belgium